Green roofs are a proven measure to increase evapotranspiration at the expense of runoff, thus complementing contemporary stormwater management efforts to minimize pluvial flooding in cities. This effect has been quantified by numerous studies, ranging from experimental field campaigns to modeling experiments and even combinations of both. However, up until now, most green roof studies consider standard types of green roof dimensions, thus neglecting varying flow length in the substrate. For the first time, we present a comprehensive investigation of green roofs that involves artificial rainfall experiments under laboratory conditions (42 experiments in total). We consider varying flow length and slope. The novelty lies especially in the consideration of flow lengths beyond 5 m and non-declined roofs. This experimental part is complemented by numerical modeling, employing the open-source Catchment Modeling Framework (CMF). This is set-up for Darcy and Richards flow in the green roof and calibrated utilizing a multi-objective approach, considering both runoff and hydraulic head. The results demonstrate that through maximizing flow length and minimizing slope, the runoff coefficient (i.e., percentage of rainfall that becomes runoff) for a 100 years design rainfall is significantly decreased: from ~30% to values below 10%. These findings are confirmed through numerical modeling, which proves its value in terms of achieved model skill (Kling-Gupta Efficiency ranging from 0.5 to 0.95 with a median of 0.78). Both the experimental data and the numerical model are published as open data and open-source software, respectively. Thus, this study provides new insights into green roof design with high practical relevance, whilst being reproducible.